FPTaylor

FPTaylor copied to clipboard

error bounding norm: num_of_float: inf

2

I'm trying to bound this program, which crosses two vectors and then crosses those results to get the sin of the angle between them: ``` Variables real x0 in [0,1.0000000000000004];...

smcallis

Does not build with ocamlc >= 4.06

1

If you compile with e.g. ocamlc 4.06 you get a bunch of warnings like this: ``` If you need to stay compatible with OCaml < 4.07, you can use the...

sampollard

arcsine(X) when X=[-1, 1]

1

I have been playing with the **arcsine** operator in FPTaylor. FPTaylor does not like when the argument of **arcsine** is a fresh variable bounded in [-1, 1]. Program causing the...

rospoly

Hexadecimal floating-point numerals

1

Use hexadecimal floating-point numerals for external optimization tools. Use `"%h"` in OCaml `printf` functions. OCaml >= 4.03 is required.

monadius

Add support for custom floating-point bit-widths

1

Several people asked me about this, given the recent trend in ML (and other areas) to use low, custom bit-widths for floating-points. I suspect this would not be that hard...

zvonimir

This is the branch that develops specified functions for FPTayfor

Thanhson89

Add files via upload

1

Adding comments to code and relu activator function

Thanhson89

When given a constant query on the edge of a function's valid range FPTaylor currently overapproximates the function input, thus leading to a range error. Treating constants as exact could...

IanBriggs

Commenting source code

2

@Thanhson89, who is a PhD student at Utah, has been investing a lot of time to understand the source code of FPTaylor. He is documenting his findings along the way,...

zvonimir

Add a custom error function for expressions in the form `expr +/- const`.

monadius